Pontryagin's maximum principle

Pontryagin's maximum principle is a fundamental theorem in optimal control theory, proved in 1956 by Lev Pontryagin and his students, that provides necessary conditions for an optimal control strategy. The principle states that the optimal control must maximize a quantity called the Hamiltonian — a function that combines the system dynamics, the cost functional, and auxiliary variables called costates — at every point along the trajectory. The costates evolve according to differential equations that are dual to the system dynamics, creating a coupled system in which the forward evolution of the state and the backward evolution of the costate jointly determine the optimal path.

The maximum principle is the control-theoretic analogue of the Principle of Least Action in physics. Where the action principle finds the trajectory that makes the action stationary, the maximum principle finds the control sequence that maximizes the Hamiltonian. The mathematical structure is identical: both are variational principles that optimize a functional over a space of possible paths, subject to dynamical constraints. The difference is interpretive: the action principle describes what nature does, while the maximum principle describes what an engineer or organism should do to achieve a goal. But the formal unity suggests that goal-directed behavior in artificial systems and the evolution of physical systems may share a deeper mathematical architecture — one in which optimization is not a special feature but the natural consequence of constrained dynamics.